An artificial intelligence pipeline for critical equipment thermal conditioning system design

Efficient electric machinery often needs to be accurately thermally conditioned. Heat sinks and heating surfaces frequently used to allow for precise temperature control of the critical equipment. To tackle the thermal challenges in the art, different design methodologies, such as the parametric or the topology optimization are introduced. Compared to parametric optimization, topology optimization allows for more tailored cooling solutions on elaborate geometries related to propulsion. Being based on gradient descent algorithm from the machine learning toolbox, topology optimization may suffer from local minima. In this report, the setup is designed to alleviate the risk for local minima and instead aim for a more global optimization. Accordingly, an artificial intelligence pipeline is scripted to run several gradient-descent based topology optimization assessments under a genetic algorithm optimization loop. The resulting geometry is shown to substantially improve the cooling ability in the given packaging volume in a light duty battery electric vehicle with quantified reduction in CO2 emissions.